Availablemujpi connectorAgent Platforms

Agent Skills

Agent Skills define what a MUJPI agent knows how to do. Rather than hardcoding capabilities into prompts, Skills are modular, version-controlled definitions combining system instructions, authorized tools, and required schemas, allowing organizations to govern agent capabilities systematically.

Architecture & Data Flow

Skills are injected dynamically into the agent's context window and execution environment at runtime based on the requested task.

Task Initiation

Skill Resolution Engine

Context & Tool Assembly

Agent Execution

Skill-Specific Output Schema

Data Flow

Skill definitions are loaded from secure storage. They dictate how the agent formats data and which tools it invokes during execution.

Typical Use Cases

✦Assigning a "Financial Audit" skill to a compliance agent
✦Updating the "Customer Refund" skill across all support agents simultaneously
✦Testing new experimental skills in a staging environment
✦Restricting destructive IT skills to senior engineering agents

Capabilities

✦Version-controlled skill definitions
✦Skill parameterization
✦Tool bundling
✦Capability inheritance

Security Considerations

Skills encapsulate permissions; possessing a skill grants the agent the authorization to use its associated tools.
Modifying production skills requires administrative approval.
Skills handling PII must be flagged to trigger compliance logging.

Setup & Deployment

Deployment Modes

Managed via MUJPI Control CenterDeployed via GitOps infrastructure as code

Setup Overview

Define a skill in JSON or via the UI, linking it to specific MCP servers or REST APIs. Assign the skill to an agent profile.